Published ahead of print on September 5, 2008, doi:10.1164/rccm.200804-494OC
© 2008 American Thoracic Society doi: 10.1164/rccm.200804-494OC
The Safety of Long-Acting β-Agonists among Patients with Asthma Using Inhaled CorticosteroidsSystematic Review and Metaanalysis niak2,3 ek2,3,4
1 Department of Medicine, McMaster University, Hamilton, Ontario, Canada; 2 Polish Institute of Evidence-Based Medicine, Krakow, Poland; 3 II Katedra Chorób Wewn Correspondence and requests for reprints should be addressed to Roman Jaeschke, M.D., 301 James Street South, Fontbonne Building, Room F506, Hamilton, ON, L8P 3B6 Canada. E-mail: jaeschke{at}mcmaster.ca
Rationale: Inhaled long-acting β-agonists (LABAs), when used as monotherapy in asthma, may increase asthma-related hospitalizations, life threatening events requiring intubation/mechanical ventilation, and asthma-related deaths, but concomitant use of inhaled corticosteroids (ICS) may modify this effect. Objectives: To determine the safety of long-acting β-agonists among patients with asthma using corticosteroids. Methods: We conducted a systematic review and metaanalysis of parallel-group, blinded, randomized, controlled trials with at least 12 weeks of treatment addressing the impact of LABA on asthma-related and total morbidity and mortality in patients concomitantly using ICS. We searched MEDLINE, EMBASE, ACPJC, and Cochrane (Central) databases, and contacted authors and sponsors.
Measurements and Main Results: We used a random effects model to pool results from different studies as odds ratios (ORs) (95% confidence interval [CI]) (OR < 1.0 favors LABA). The search yielded 62 relevant studies included in this analysis. Among over 29,000 participants (15,710 taking LABA, with over 8,000 patient-years observed in the LABA groups), there were three asthma-related deaths and two asthma-related, nonfatal intubations (all in LABA groups; Conclusions: In patients with asthma using ICS, LABA did not increase the risk of asthma-related hospitalizations. There were very few asthma-related deaths and intubations, and events were too infrequent to establish LABA's relative effect on these outcomes.
Key Words: long-acting β-agonists • randomized trials • toxicity • adverse events • systematic review
Harmful effects of long-acting β-agonists (LABAs) in patients with asthma have been suggested by a large, randomized trial of salmeterol versus salbutamol by Castle and colleagues (1); more recently by another trial of salmeterol versus placebo added to usual therapy by Nelson and colleagues (2); and finally by a systematic review of data for both salmeterol and formoterol performed by Salpeter and colleagues (3), but heavily influenced by the study by Nelson and colleagues. Harmful effects have included severe asthma exacerbations requiring hospitalization, life-threatening exacerbations requiring intubations, and asthma-related death (4). These reports, however, largely reflect data from clinical settings in which inhaled steroids were not mandated as a background treatment. Asthma management guidelines have consistently recommended that LABAs should only be used in combination with inhaled corticosteroids (ICS) (5, 6). Systematic reviews have not yet comprehensively addressed the impact of LABAs on these most serious outcomes in the presence of mandated ICS in populations receiving, and not receiving, LABAs (7–10). We therefore conducted a systematic review to examine the safety of LABAs (both formoterol and salmeterol) when taken regularly by patients with asthma who are also taking ICS. Specifically, we address the relative and absolute rate of serious adverse events (SAEs) in randomized trials of patients treated with LABAs versus control populations not receiving LABAs. Outcomes of interest include death, life-threatening situations leading to intubation and ventilation, hospitalization, and SAEs (in each case, both asthma-related and total). Although the primary focus was LABA safety in studies in which the dose of ICS was similar among patients receiving and not receiving LABA, we also evaluated studies using higher doses of ICS among patients not receiving LABA. Some of the results of this study have been previously reported in abstracts (11, 12).
Eligibility Criteria We included studies with the following characteristics: treatment allocation by randomization; parallel control groups (crossover studies excluded) with at least 12 weeks of treatment; blinding of patients and care-givers; acceptable follow-up of patients receiving study medication (outcome data for the full duration of planned treatment missing for <20% of patients taking LABA in trials  3 mo long, <30% for 3 mo to <1 yr long, <40% for  1 yr). Eligible studies involved patients with asthma (excluding children younger than 12 yr); all patients had to be receiving at least some ICS (in studies having more than 2 treatment groups, we included patients in groups mandated to receive some ICS).
Intervention Group
Control Group
Outcome Measures The secondary prespecified outcomes were: total nonfatal intubation and ventilation; total nonfatal hospitalization; total nonfatal SAE; asthma-related death or intubation; and total death or intubation. We accepted the authors' or manufacturers' classifications of events as asthma-related or non–asthma-related.
Search Strategy—Identification of Studies Two reviewers (R.J. and J.B., F.M., P.N. or W.L.) screened titles and abstract to identify articles for full review. These same reviewers then evaluated the full text of all articles deemed potentially eligible by either reviewer for final inclusion.
Data Extraction
Statistical Analysis In choosing the appropriateness of combining data, we considered the following a priori rules for the interpretation of heterogeneity of results (measured by I2): 0–30% as low; 30–80% as moderate and worthy of investigation; 60–100% as large and worthy of understanding; 95–100% do not aggregate. In addition, we explored heterogeneity by performing univariable and multivariable metaregressions to investigate the following a priori hypotheses of factors potentially influencing effect size: medication used (salmeterol vs. formoterol); dose of ICS used in the control group (similar dose of ICS to LABA group or increased dose); and use of ICS as part of study medications in single device with LABA, in separate study devices, or as a nonstudy background medication. These analyses were performed using STATA version 9.2 (StataCorp LP, College Station, TX).
In the univariable analysis, the criterion for statistical significance for each of the above analyses, performed using a t test, was set a priori at an For the estimation of number of patient-years on LABA treatment, when explicit data were not available, we assumed that patients who did not complete the total duration of follow-up for a given study received medication assigned by randomization for 50% of the study duration (i.e., for 12 wk in 24-wk studies).
Of 1,402 titles and abstracts identified during the primary search, and an additional 17 identified by manufacturers of the drugs under investigation, we obtained the full manuscript of 234 publications (Figure 1). Of those, 63 publications describing 65 studies were eligible (20 publications involving formoterol and 43 publications describing 45 studies involving salmeterol).
The reasons for exclusions of studies from the analysis were in most cases multiple, but included: not all participating patients received ICS (37 articles); review article/editorial/letter/protocol rather than original article (26 articles); use of data for multiple publication from one study (usually in the form of additional economic analysis, quality of life data, etc.) (24 articles); crossover design (18 articles); duration less than 12 weeks (16 articles); pediatric population (16 articles); use of LABA in both groups (10 articles); lack of blinding of patients or care-givers (seven articles); excessive missing data for patients receiving study medication (six articles); inclusion of patients with chronic obstructive pulmonary disease (three articles); not using LABA in all patients in one group (three articles); and comparison of LABA with SABA (one article). In addition, we were unable to obtain full data from five sources and data dealing with one subgroup from one source: no final response to the request for full data (16–18); one study finished, but not yet published, either in a peer-reviewed journal or on the Clinical Trial Register maintained by the manufacturer; in two cases, we were unable to obtain full sets of data from groups of patients using 100 µg of salmeterol twice a day (19, 20). Of note, in the primary articles describing the results of those studies (16–20), there was no record of deaths or intubations. Only one additional publication was identified during an update of the manuscript in April 2008 (2007 Corren, see online supplement). Thus, we have analyzed data from 62 studies described in 60 publications. These publications are listed in the online supplement, as are the details about included studies. We found no case of discrepancy in number of deaths or intubations reported among the published data compared with information from sponsors or information from authors. All discrepancies concerning frequency of hospitalizations or SAE involved no more than one event per study group and/or one patient in the denominator for a given group.
Studies Using Formoterol
Formoterol (other sponsors and manufacturers).
Studies Using Salmeterol
Effects of Treatment—Salmeterol and Formoterol Studies Combined
Exploration of Heterogeneity Individual study results were largely consistent for all outcomes (tests for heterogeneity while using RevMan had P values not lower than 0.30; I2 < 10%). Examination of the a priori hypotheses to explore the heterogeneity that was present (type of drug, similarity of ICS dose in the LABA and control groups, and the mode of drug delivery investigated, each for several outcomes) failed, in almost all cases, to explain the small degree of heterogeneity that existed. In only two cases did we find P values close to the prespecified threshold of 0.1. First, for asthma-related hospitalization (Figure 2), the univariable analysis for the test of interaction related to the two formulations (i.e., testing the hypothesis that the effect of formoterol differs from the effect of salmeterol) resulted in a P value of 0.13. This result raises the possibility that formoterol may yield more favorable effects than salmeterol on asthma-related hospitalization. Second, for asthma-related SAEs, the univariable analysis for the test of interaction related to the two formulations resulted in a P value of 0.06, again raising the possibility of a more favorable effect of formoterol than salmeterol. In multivariable analyses, the P value for the effect of drug was 0.07 for asthma-related SAE and 0.12 for asthma-related hospitalizations. Asthma-related deaths (two deaths in studies using formoterol and one in studies using salmeterol) and asthma-related intubation and ventilation (none and two, respectively) were too infrequent to analyze. For total mortality, the P value for the test of interaction related to formulation was 0.82.
This metaanalysis of the effect of LABAs in combination with ICS on serious, harmful effects in clinical trials of asthma did not show an increased risk for hospitalizations or SAEs, whereas the relative effect on asthma-related mortality and asthma-related intubation and ventilation could not be assessed, because of the very low frequency of these events. This contrasts with the report of Nelson and colleagues (2), who reported increased risk of respiratory-related death (relative risk [RR], 2.16; 95% CI, 1.06–4.41), asthma-related deaths (RR, 4.37; 95% CI, 1.25–15.34), and of combined asthma-related deaths and life-threatening experiences (RR, 1.71; 95% CI, 1.01–2.89) associated with the use of salmeterol. A subsequent systematic review and metaanalysis largely replicated those findings, and also recorded significant increase in the risk of asthma-related hospitalization (OR, 2.6; 95% CI, 1.6–4.3) (3). Considerations regarding the potential mechanisms of these adverse effects included masking of increasing inflammation, delaying awareness of worsening asthma (21), and induction of bronchodilator subsensitivity (22, 23). Our study did not confirm those findings in a setting mandating use of ICS. The studies that found increased deaths and hospitalizations with use of LABAs did not mandate ICS use, and included many patients who were not taking ICSs. Although they are effective and long-lasting bronchodilators, LABAs are not known to have clinically significant antiinflammatory effects in the airways. No recently published asthma treatment guidelines recommend LABA use without ICS (5, 6). Hence, we have considered a different population: patients who were receiving ICS concomitantly with LABAs. Indeed, there is only a single study (2002 Price, see online supplement) common to the 19 studies in the previous publication by Salpeter (3) and the 62 studies in the current metaanalysis. The differences in the results from previous metaanalysis are thus not entirely unexpected, but nevertheless require exploration.
Decreased LABA exposure cannot explain the substantially smaller number of asthma-related deaths and intubations experienced by patients in the studies that we examined compared with the previous metaanalyses, as the total number of patient-years exposure to LABA was similar (
In our analysis of trials in which all patients used ICS, we did not find a significant increase in deaths or life-threatening events. Our results do not, however, exclude a relative increase in deaths—in particular, asthma-related deaths—and asthma-related intubations associated with LABA use, because patients receiving LABAs experienced the few asthma-related deaths and intubations that did occur. Our results show that the absolute increase in LABA-associated deaths or intubations from asthma in populations, such as those participating in these trials, is small, if it exists at all (3 deaths and 2 nonfatal intubations in 15,710 patients receiving LABA). The frequency of asthma-related intubations appears approximately 10-fold lower among patients included in the current metaanalysis when compared with that of Salpeter and colleagues (3). Furthermore, our results exclude an important increase in asthma-related hospitalizations or SAEs, and suggest the possibility of benefit of LABA used with ICS on these outcomes. Our findings contrast with the recent Cochrane Review of 62,630 patients in trials comparing salmeterol with placebo or salbutamol, in which ICS use was not required in all patients (10). In those studies, there was a significant increase in nonfatal SAEs related to use of salmeterol (OR, 1.14; 95% CI, 1.01–1.28). Less severe underlying asthma in patients enrolled in LABA comparison studies in which ICS was mandated when compared with those in which ICS was not mandated could explain the difference in results. Nevertheless, such an explanation seems unlikely: The mean age of patients enrolled in the study by Nelson and colleagues (2) was 39.1 years (in the midrange for the studies included in our metaanalysis), and the mean PEF was 83.9% of predicted; the main inclusion criteria of Nelson and colleagues were clinical diagnosis of asthma and receiving asthma prescription medication, whereas the majority of studies in our metaanalysis required patients to be symptomatic despite treatment with ICS. All these factors make it unlikely that underlying asthma severity was markedly higher in the study population of Nelson and colleagues, and was thus a cause of different results. A third possibility is that other differences between the randomized trials that we examined (phase-3 studies) and the studies suggesting an increased mortality with LABA (postmarketing or phase-4 studies) may explain the varying results. Such factors might include less careful management and more overuse of LABA and SABA in the less rigorously controlled trials. One cannot disprove this hypothesis with certainty, although the accompanying observation of lower number of asthma-related hospitalizations speaks against it. Finally, one may postulate that the prognosis of asthma and the potential effects of LABA taken regularly are different among patients taking and those not taking concomitant ICS. This hypothesis suggests that ICS provides both protection against severe asthma deterioration and protection against the potential harmful actions of LABA. Our analysis, however, had too few of the most severe events to make definitive statements about the relative effects on deaths and intubations. The effect, if it exists at all, is smaller than can be inferred from the analysis of Salpeter and colleagues (3). On the other hand, given the number of patients using LABA worldwide, any true underlying increase in mortality would be important. Thus, among patients with asthma that is not well controlled with low doses of ICS, those concerned about the remaining uncertainty (and their physicians) may well prefer increasing the dose of ICS instead of adding LABA. Our exploration of heterogeneity suggests that formoterol may result in fewer asthma-related hospitalizations and asthma-related SAEs than salmeterol. The hypothesis was one of only three a priori hypotheses that we examined, and the magnitude of the difference was relatively large (Figure 2). These factors support an inference of a real difference between drugs. On the other hand, the studies represent indirect rather than head-to-head comparison, and were conducted in different time periods and locations. In addition, patients did not experience hospitalizations in different proportion of studies (in over 50% of studies using salmeterol, and under 25% of studies using formoterol). Furthermore, the P value of the test for interaction is unimpressive (0.12 for asthma-related hospitalizations). Thus, the inference that the drugs have different effects on hospitalization is weak, and only a direct head-to-head comparison may establish whether there is a true difference (24, 25). Similar reasoning applies to the indirect comparisons of SAEs. The added value of our analysis in comparison with related clinical questions examined by others (8–10, 26, 27) relates to: focusing on specific questions not examined before in detail; a larger set of high-quality studies for which data were sought and obtained through contact with both authors (some studies) and manufacturers (all studies); obtaining and including complete data on mortality and respiratory failure requiring intubation, both asthma-related and total; and cooperation of the main manufacturers of LABA (AstraZeneca and GlaxoSmithKline) with disclosure of their data. The main limitation of our study is that we were unable to fully explore alternative explanations of the difference in death and intubation rates in studies mandating concomitant ICS use and those that did not. Exploration of factors such as socioeconomic status or race would require both measurement of those variables and access to individual patient data in studies mandating and not mandating concomitant ICS use. Additional limitations are our exclusion of children less than 12 years of age, and our failure to explore similarities or differences in effect in this subpopulation. In summary, our systematic review evaluated the safety of adding LABA to ICS, and did not reveal convincing evidence of harm. We have extended previous reports that addition of LABA has no detrimental effect on the number of asthma-related hospitalizations and SAEs in patients receiving ICS (8, 9). We have not, however, excluded the possibility of a relative increase in deaths in patients receiving LABA who are also using ICS, a possible increase that may be important at a population level or to individual patients. The extent to which our results are reassuring regarding the use of LABA in patients receiving ICS may differ according to the perspective of physician, patient, and policy maker.
This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org Originally Published in Press as DOI: 10.1164/rccm.200804-494OC on September 5, 2008 Conflict of Interest Statement: R.J. received honoraria for lectures from Astra-Zeneca, Merck Sharp & Dohme, Boehringer Ingelheim, and GlaxoSmithKline (GSK), and travel support from Boehringer Ingelheim and GSK, and is a deputy editor of a medical journal (Medcyna Praktyczna) funded in part by industry advertisement, including manufacturers of the drugs that are the subject of this review. P.M.O. has been on advisory boards for AstraZeneca, Biolipox, GSK, Nycomed, Ono Pharma, Topingen, and Resistencia, and has received lecture fees from these pharmaceutical companies, as well as from Chiesi, and holds industry-sponsored grants from AstraZeneca, Altana, Chiesi, Biolipox, Boehringer, GSK, Medimmune, Merck, Pfizer, and Wyeth. F.M. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. P.N. holds investigator-initiated grants from GSK Canada, Sepracor Inc., and Topigen Inc., and two investigator-initiated grants from a Firestone-Pharmaceutical Industry Strategic Collaboration. He received honoraria from Merck Frosst, AstraZeneca, GSK, and Altana Pharma, and his research group has a patent on a sputum filtration device. W.L. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. J.B. is a journal editor and receives a salary from a clinical journal (Medycyna Praktyczna) in which various drugs are advertised, including those that are the subject of this systematic review. J.B. received an honorarium from GSK. L.T. will be conducting consulting work for GSK in the near future. J.C. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. H.J.S. received research grants and fees and/or honoraria from AstraZeneca, Chiesi Foundation, Lily, Pfizer, Roche, and UnitedBioSource for development or consulting regarding quality-of-life instruments for chronic respiratory diseases, or as lecture fees related to the methodology of evidence-based practice guideline development and/or research methodology. M.R.S. has received research funding and/or an honorarium for lectures and consultations from AstraZeneca, GSK, Merck, Nycomed, and Schering Plough, received grants over the last 5 years from AstraZeneca, Boehringer Ingelheim, GSK, and Merck Frosst Canada, and holds a chair in respiratory epidemiology jointly endowed by AstraZeneca and McMaster University. G.G. received funding from the following pharmaceutical companies to carry out research projects: Lotte and John Hecht Foundation, Bristol-Myers Squibb, and Pfizer Inc., and is currently acting as a consultant and received honoraria from Eli Lilly Canada Inc. Received in original form April 1, 2008; accepted in final form September 4, 2008
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
trznych, Jagiellonian University School of Medicine, Krakow, Poland; 4 Italian National Cancer Institute, Rome, Italy; 5 Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, Canada; 6 Biostatistics Unit, Father Sean O'Sullivan Research Centre, St. Joseph's Healthcare, Hamilton, Ontario, Canada; and 7 CLARITY Research Group, Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, Canada
TOP
ABSTRACT
AT A GLANCE COMMENTARY
of 0.10. We chose 0.10 as a threshold P value because we anticipated limited power to this analysis, and did not want to overlook possible explanations of heterogeneity that warranted serious consideration. 
8,000 patient-years in both data sets). 
